[Kernel Bug] Kernel 3.15 broke my radeon graphics!

rabcor · 2014-06-19 13:36:00

I ran an update on my laptop like yesterday, when I launched it today I ended up with a frozen display manager. My first thought was broken display manager but turns out it was X. I have no idea what's up, I'm using the radeon open source drivers with a hybrid graphics card, another noteable problem that came up is that "/sys/kernel/debug/vgaswitcheroo/switch" (file I used to disable these hybrid graphics) does not exist anymore either.

Here is my xorg log but I suspect the problem is significantly deeper than just xorg failing (it is vital for me to disable the hybrid graphics as it runs hotter than any other component in the laptop) also on bootup it is giving me errors:

pciehp 0000:00:02.0:pcie04: slot(2): Link Up event
pciehp 0000:00:02.0:pcie04: Device 0000:01:00.0 already exists at 0000:01:00, cannot hot-add
pciehp 0000:00:02.0:pcie04: Cannot add device at 0000:01:00
pciehp 0000:00:02.0:pcie04: Card not present on Slot(2)
pciehp 0000:00:02.0:pcie04: slot(2): Link Down event
pciehp 0000:00:02.0:pcie04: Link Down event ignored on slot(2): already powering off

It seems my hybrid graphics thing broke might be a kernel problem with 3.15 and my hp pavilion dv7-6109eo?

Update: Yes it was a kernel problem. kernel 3.15 seems to have broken compatibility with the radeon hybrid graphics card on hp pavilion dv7-6109eo. I downgraded to 3.14-6 and it works fine again. Edited title too to match the problem. Perhaps this bug should be reported somewhere... In the meantime, a method to prevent system updates from updating the kernel would be nice.

Update: adding this to the commandline allows me to start X without everything freezing up. But I still get all of the other issues (pciehp cannot hot-add and /sys/kernel/debug/vgaswitcheroo does not exist)

radeon.runpm=0

Normal pciehp output from journalctl on kernel 3.14.6:

pciehp 0000:00:02.0:pcie04: Slot #2 AttnBtn- AttnInd- PwrInd- PwrCtrl- MRL- Interlock- NoCompl+ LLActRep+
pciehp 0000:00:02.0:pcie04: service driver pciehp loaded
pciehp: PCI Express Hot Plug Controller Driver version: 0.4
pciehp 0000:00:02.0:pcie04: Card not present on Slot(2)

Last edited by rabcor (2014-06-25 22:38:01)

LQYMGT · 2014-06-20 02:38:55

well, I meet this problem too...

Pse · 2014-06-20 03:46:28

You can use the --ignore linux pacman argument to prevent updates to the kernel. Be aware that this is only a temporary solution. You could also switch to the LTS branch of the linux kernel by installing linux-lts.

zwa · 2014-06-20 05:04:28

I have a Radeon HD 4350 running with the oss drivers and linux 3.15, I have not experienced this problem.

LQYMGT · 2014-06-20 13:58:17

LQYMGT wrote:

well, I meet this problem too...

Ennn...I got it.
Try to add radeon.runpm=0 to your kernel parameters.

rabcor · 2014-06-25 22:07:34

zwa wrote:

I have a Radeon HD 4350 running with the oss drivers and linux 3.15, I have not experienced this problem.

It is associated with switchable graphics cards. I think HP actually modifies the radeon GPUs in their systems somehow (it is not the same card as AMD officially manufactures, and this has gotten me into some problems, more on Linux than on Windows, but i have also run into problems on windows)

LQYMGT wrote:

add radeon.runpm=0 to your kernel parameters

This worked! thanks, but I still get

pciehp 0000:00:02.0:pcie04: slot(2): Link Up event
pciehp 0000:00:02.0:pcie04: Device 0000:01:00.0 already exists at 0000:01:00, cannot hot-add
pciehp 0000:00:02.0:pcie04: Cannot add device at 0000:01:00
pciehp 0000:00:02.0:pcie04: Card not present on Slot(2)
pciehp 0000:00:02.0:pcie04: slot(2): Link Down event
pciehp 0000:00:02.0:pcie04: Link Down event ignored on slot(2): already powering off

and "/sys/kernel/debug/vgaswitcheroo" does not exist either. (It is crucial for me to be able to disable the switchable graphics and just use the intergrated card rather than the dedicated one)

0000:00:02:0 (lspci) is

00:02.0 PCI bridge: Advanced Micro Devices, Inc. [AMD] Family 12h Processor Root Port (prog-if 00 [Normal decode])

Using kernel module shpchp and driver pcieport.

Normal reaction coming from kernel 3.14.6:

pciehp 0000:00:02.0:pcie04: Slot #2 AttnBtn- AttnInd- PwrInd- PwrCtrl- MRL- Interlock- NoCompl+ LLActRep+
pciehp 0000:00:02.0:pcie04: service driver pciehp loaded
pciehp: PCI Express Hot Plug Controller Driver version: 0.4
pciehp 0000:00:02.0:pcie04: Card not present on Slot(2)

Last edited by rabcor (2014-06-25 22:30:34)

dobrev92 · 2014-06-30 19:04:59

Situation confirmed on pavilion g6.
Ironic, considering that I read somewhere that this kernel offered vast improvements for Radeon cards

lbdkmjdf · 2014-07-08 21:53:55

Guys, I need your help to report and fix this.
I reported this bug: https://bugzilla.kernel.org/show_bug.cgi?id=79701 It would be great if you guys do the testing/bisecting/etc. and report feedback.
There is a similar thread in this forum: https://bbs.archlinux.org/viewtopic.php?pid=1433713 , it would be cool if mods mixed both threads.

clfarron4 · 2014-07-08 22:23:23

@rabcor, have you actually tried running linux-lts (which is now on the 3.14 branch)?

rabcor · 2014-07-12 09:58:51

No but I'm sure this would work as it is not 3.15 (i.e. graphics broke in 3.15)

I also notice (for the better in my case) that the system when I launch it on 3.15 kernel with radeon.runpm=0 it will use the intergrated card rather than the discrete one (it fails to detect the discrete one which may be the real problem)

Running LTS would only be delaying the inevitable (but thanks for the pointer, my laptop might actually be better off on LTS so I might switch to it after all; when this issue has been solved). Merging the threads is probably not needed (only the OP of the other post is relevant, it would have needed to be merged sooner)

I noticed (probably related to this issue) that my computer still sometimes fails to launch graphics on startup and when it occurssystemd gives an error that it cannot finalize shutting down "remaining file systems and devices, giving up" then it fails to unmonut /oldroot and subirectories (systemd gives up on restarting)

Pressing the shutdown button in this state gives

atkbd serio0: Unknown key pressed (translated set 2 , code 0xab on isa0060/serio0).
atkbd serio0: Use 'setkeycodes e02b <keycode>' to make it known.

Only way to finish shutdown is forcing it down by holding down the shutdown key.

Here is a journalctl -b to provide insight. Lines to look at are 545-549, 591-592, 706, 817, 952-1198. It is worth noting that all of the errors on these specific lines occur both on a successful boot (where I get into mdm) and on the boot that fails (even with radeon.runpm=0 set)

This event only seems to occur on cold boots (and by cold I don't just mean in the technical sense of the word, but also when the computer actually is physically cold, as in there are hardware temperature implications that contribute to this problem, but this is of course also a kernel issue since it never happened on 3.14)

Last edited by rabcor (2014-07-12 10:15:08)

bijanbina · 2014-07-13 13:57:50

I have the same problem ,after upgrading i cant use radeon any more(just intel)

then i take these steps in order to re enable vgaswitcheroo
1.delete radeon from black list which i found @ /etc/modprobe.d
2.change /etc/mkinitcpio.conf MODULES to "radeon"
3.add mentioned kernel parameter

after reboot vgaswitcheroo enabled so i could see

$ sudo cat /sys/kernel/debug/vgaswitcheroo/switch
0:IGD:+:Pwr:0000:00:02.0
1:DIS: :Pwr:0000:01:00.0
2:DIS-Audio: :Pwr:0000:01:00.1

but my problem is that i cant turn ati GPU off neither befor do that nor after

when radeon is enabled my laptop is getting hot and hotter so i want to disable it

i can do that in 3.14 but not in new one

note:my laptop is MUX hybrid one!

Last edited by bijanbina (2014-07-13 13:59:47)

lbdkmjdf · 2014-07-13 23:55:17

Anyone cares about the bug report?
Please, add yourself to the report, so we can add pressure for devs to fix it: https://bugzilla.kernel.org/show_bug.cgi?id=79701

lbdkmjdf · 2014-07-15 02:46:03

Update: As shown in https://bugzilla.kernel.org/show_bug.cgi?id=79701#c2 and http://www.phoronix.com/forums/showthre … post428972 , the solution (or, perhaps, workaround) is to disable "OSC" from the BIOS.

rabcor · 2014-07-26 21:59:48

There is no OSC option anywhere in my BIOS settings, to be honest my BIOS settings are the most minimal I have ever seen on any computer. I can probably count all the options on just my fingers.

Arachnid92 · 2014-07-27 22:39:31

rabcor wrote:

There is no OSC option anywhere in my BIOS settings, to be honest my BIOS settings are the most minimal I have ever seen on any computer. I can probably count all the options on just my fingers.

I don't have this option in the BIOS either, and I'm experiencing similar issues (discrete card disappeared ).

lbdkmjdf · 2014-07-31 00:23:35

Looks like the bug is being worked at: https://bugzilla.kernel.org/show_activity.cgi?id=79701 . Is anyone running a 32 bits kernel in a dual AMD graphics system?

rabcor wrote:

There is no OSC option anywhere in my BIOS settings, to be honest my BIOS settings are the most minimal I have ever seen on any computer. I can probably count all the options on just my fingers.

Yeah, I'm running a modified BIOS, and that's why I can see/switch that setting... disabling runpm and using vgaswitcheroo is the only workaround I know of.

Arachnid92 wrote:

rabcor wrote:
There is no OSC option anywhere in my BIOS settings, to be honest my BIOS settings are the most minimal I have ever seen on any computer. I can probably count all the options on just my fingers.
I don't have this option in the BIOS either, and I'm experiencing similar issues (discrete card disappeared ).

I think the discrete card "dissapears" *because* runpm is working... to use it, you have to run something like:

$ DRI_PRIME=1 ./some_program

If you add 'radeon.runpm=0' in the kernel command line, you should be able to see your dedicated card, but you would have to enable/disable it manually (look for a tutorial on vgaswitcheroo). This is not the same issue as in the OP...

Last edited by lbdkmjdf (2014-07-31 00:28:22)

Arch Linux

#1 2014-06-19 13:36:00

[Kernel Bug] Kernel 3.15 broke my radeon graphics!

#2 2014-06-20 02:38:55

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#3 2014-06-20 03:46:28

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#4 2014-06-20 05:04:28

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#5 2014-06-20 13:58:17

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#6 2014-06-25 22:07:34

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#7 2014-06-30 19:04:59

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#8 2014-07-08 21:53:55

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#9 2014-07-08 22:23:23

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#10 2014-07-12 09:58:51

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#11 2014-07-13 13:57:50

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#12 2014-07-13 23:55:17

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#13 2014-07-15 02:46:03

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#14 2014-07-26 21:59:48

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#15 2014-07-27 22:39:31

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

#16 2014-07-31 00:23:35

Re: [Kernel Bug] Kernel 3.15 broke my radeon graphics!

Board footer